You are on page 1of 3

Asrar Chowdhury

GT Handout 8: Zero Sum Games and Saddle Points

Zero-Sum Games and Saddle Points


B: North South

A: East West

Row 1

Col 3
1

Likes High

Row 2

Row Min: 2

3
Saddle Point
Nash equilibrium

Row Min: 3

Col Max: 7

Col Max: 3

Col Max: 5

Altitudes

Col 1

Col 2

Likes Low
Altitudes

Row 3

Minimax: min
from set of maxima

Row Min: 1

B chooses min from As


chosen set of maxima

Maximin: max
from set of
Minima
A chooses max
from Bs chosen set
of minima

If a zero-sum game has a saddle point, a Nash equilibrium can be found in pure
strategies.
A saddle point is a point in a domain of a function that is a stationary point, but not a
local extremum. A saddle point is a coordinate that is a maximum in one direction, and
a minimum in another direction. A saddle point occurs in a zero-sum game under the
following condition.
A saddle point exists when the maximin outcome for a row player using a pure strategy
coincides with the minimax outcome for a column player using a pure strategy. A
saddle point is also a Nash equilibrium- they are mutual best responses against each
other. Once a saddle point is reached, no player in the game will want to unilaterally
deviate from it. There are two types of strategies the players will be using in the
presence of a Saddle Point. They are:
Maximin Strategy: The best (highest) payoff from a set of minima. Player A- the Row
Player will opt for this strategy. Minimax Strategy: The worst (lowest) payoff from a
set of maxima. Player B- the Column Player will opt for this strategy.
Two friends A and B have to decide on a holiday location for camping. Friend A is the
row player who chooses a site on the East-West track. Friend A likes high altitudes and
would ideally choose a site as high as possible. Friend B is the column player who
chooses a site on the North-South track. Friend B does not like high altitudes and would
ideally choose a site as near as possible to sea level.
The two friends have mutually conflicting interests. However, unlike the Matching
Pennies Game, the game is not non-cooperative. The two players have to establish a
binding contract if they want to enjoy the vacation together. The payoffs are shown in
the matrix above. The payoffs are shown as the heights of the holiday location. The
heights can be viewed as a zero-sum game. Each height will bring comfort (positive
2013 GT: 01819 219050 asrarul@gmail.com

-1

asrar.chowdhury@facebook.com

Asrar Chowdhury

GT Handout 8: Zero Sum Games and Saddle Points

utility) to one player, but discomfort (negative utility) to the other player. The features
of the game:
The game is a zero-sum game. Player A and B have mutually conflicting interests. Player A
likes high altitudes; and Player B likes low altitudes. Since A and B want to spend a vacation
together they need to establish a binding contract. The game needs to be a cooperative. The Nash
equilibrium will need both players to play a conservative strategy to reach equilibrium. This
means they will play maximin and minimax strategies.
RESPONSES OF PLAYER A to Bs Chosen Actions:
Ideally Player A would choose high altitudes. Since Player B chooses low altitudes,
Player A chooses the minimum of each Row. Player As choice depends on what Player
B will do. Player B does not like heights and will thus choose low heights. This is
consistent with inter-dependent decision-making of Game Theory.
This results in the following set of minima from Row 1-3: (1,2,3). The maximin strategy
then says: Player A chooses the maximum from this set of minima. This is 3- the
maximum from the three row minima in Row 3.
RESPONSES OF PLAYER B to As Chosen Actions:
Ideally Player B would choose low altitudes. Since Player A chooses high altitudes,
Player B chooses the maximum of each Row. Player Bs choice depends on what Player
A will do. Player A likes heights and will thus choose high heights. This is consistent
with inter-dependent decision-making of Game Theory.
This results in the following set of maxima from Row 1-3: (7,3,5). The minimax strategy
then says: Player B chooses the minimum from this set of maxima. This is 3- the
minimum from the three column maxima in Column 2.
The intersection of the maximin and the minimax gives the Saddle Point. This arises at
the intersection of Row 3 and Column 2. The maximin and the minimax strategies arise
because of mutually conflicting interests. Since the game is cooperative the following is
also observed:
1. The payoffs are not symmetrical. The Row Minima and the Column Maxima are
different. Row Minima: (1,2,3); Column Maxima: (7,3,5).
2. When Row Minima and Column Maxima are not the same a Saddle Point exists
in a zero-sum game. Players can establish a binding contract. A cooperative game
can be played.

2013 GT: 01819 219050 asrarul@gmail.com

-2

asrar.chowdhury@facebook.com

Asrar Chowdhury

GT Handout 8: Zero Sum Games and Saddle Points

The Saddle Point- at the intersection of the maximin and minimax- is also a Nash
equilibrium. The reasons are as follows:
1. The Saddle Point is based on the maximin and minimax strategies where the
players are choosing their best responses in accordance to the behaviour of the
other player. The behaviour is conservative because the maximising player
(Row, A) is choosing the maximum payoff from a set of minima (the minimum
values determined by the other player); and the minimising player (Column, B)
is choosing the minimum payoff from a set of maxima (the maximum value
determined by the other player).
2. The intersection of maximin and the minimax gives the Saddle Point which is a
mutual best response of both players given what the other player will do. It also
qualifies as a Nash equilibrium. Once a saddle point is reached at, no player will
unilaterally want to deviate. Deviation along the Row to 5 or 4 is desirable for
Player A, but not desirable for Player B. Deviation along the Column to 2 is
desirable for Player B, but for Player A. The Saddle Point is thus a stable
equilibrium and also a Nash equilibrium.
3. The conservative attitude of the players can also be explained observing the
range of the heights in this example that ranges from 1 to 7. Ideally, Player A
would want to be at the right extreme: 7, but moves downwards because Player
wants to move in the opposite direction. Conversely, Player B would like to be at
the left extreme: 1, but moves upwards because Player A wants to move in the
opposite direction. Their mutually conflicting interest in a cooperative game
leads them to rest somewhere in the middle at 3 (not necessarily at the centre, but
around it).
A zero-sum game has a Saddle Point only if it is cooperative. The Matching Pennies
Game does not have a Saddle Point because it is non-cooperative. Players seek a
conservative strategy when a Saddle Point exists and thus the Maxi-Min and Mini-Max
strategies. Non cooperation in the Matching Pennies game indicates players will have
aggressive strategies. Therefore a Maxi-Min and Mini-Max strategy will not exist.
John von Neumann- The Father of Game Theory- introduced the Saddle Point solution
to solve zero-sum games in 1928. This is also the same year John Nash was born. The
solution came from a conjecture from the French mathematician Emile Borel that zerosum games do not have a solution. The Saddle Point is the solution since the concept of
the Nash equilibrium did not appear until 1950.
-- The End -

Adapted from: Complete Idiots Guide to Game Theory: The fascinating math behind
decision-making. Edward C Rosenthal. New York: Penguin. 2011.
2013 GT: 01819 219050 asrarul@gmail.com

-3

asrar.chowdhury@facebook.com

You might also like